From: James Smart <jsmart2021@gmail.com>
Date: Mon, 9 Apr 2018 14:24:22 -0700
Subject: scsi: lpfc: Correct target queue depth application changes
Git-commit: f91bc594ba963a9354c9da8bb85c21606c2b6289
Patch-mainline: v4.18-rc1
References: bsc#1088866

The max_scsicmpl_time parameter can be used to perform scsi cmd queue
depth mgmt based on io completion time: the queue depth is reduced to
make completion time shorter. However, as soon as an io completes and
the completion time is within limits, the code immediately bumps the
queue depth limit back up to the target queue depth. Thus the procedure
restarts, effectively limiting the usefulness of adjusting queue depth
to help completion time.

This patch makes the following changes:

 - Removes the code at io completion that resets the queue depth as soon
   as within limits.

 - As the code removed was where the target queue depth was first
   applied, change target queue depth application so that it occurs when
   the parameter is changed.

 - Makes target queue depth a standard parameter: both a module
   parameter and a sysfs parameter.

 - Optimizes the command pending count by using atomics rather than
   locks.

 - Updates the debugfs nodelist stats to allow better debugging of
   pending command counts.

Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <james.smart@broadcom.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-by: Johannes Thumshirn <jthumshirn@suse.de>
---
 drivers/scsi/lpfc/lpfc.h         |  2 --
 drivers/scsi/lpfc/lpfc_attr.c    | 45 ++++++++++++++++++++++++++++++++++++++--
 drivers/scsi/lpfc/lpfc_debugfs.c | 20 ++++++++++++++++--
 drivers/scsi/lpfc/lpfc_scsi.c    | 19 +++++------------
 4 files changed, 66 insertions(+), 20 deletions(-)

diff --git a/drivers/scsi/lpfc/lpfc.h b/drivers/scsi/lpfc/lpfc.h
index 6c0d351c0d0d..be4abe52f289 100644
--- a/drivers/scsi/lpfc/lpfc.h
+++ b/drivers/scsi/lpfc/lpfc.h
@@ -64,8 +64,6 @@ struct lpfc_sli2_slim;
 #define LPFC_IOCB_LIST_CNT	2250	/* list of IOCBs for fast-path usage. */
 #define LPFC_Q_RAMP_UP_INTERVAL 120     /* lun q_depth ramp up interval */
 #define LPFC_VNAME_LEN		100	/* vport symbolic name length */
-#define LPFC_TGTQ_INTERVAL	40000	/* Min amount of time between tgt
-					   queue depth change in millisecs */
 #define LPFC_TGTQ_RAMPUP_PCENT	5	/* Target queue rampup in percentage */
 #define LPFC_MIN_TGT_QDEPTH	10
 #define LPFC_MAX_TGT_QDEPTH	0xFFFF
diff --git a/drivers/scsi/lpfc/lpfc_attr.c b/drivers/scsi/lpfc/lpfc_attr.c
index 2ac1d21c553f..c89ffad1f43d 100644
--- a/drivers/scsi/lpfc/lpfc_attr.c
+++ b/drivers/scsi/lpfc/lpfc_attr.c
@@ -3469,8 +3469,49 @@ LPFC_VPORT_ATTR_R(lun_queue_depth, 30, 1, 512,
 # tgt_queue_depth:  This parameter is used to limit the number of outstanding
 # commands per target port. Value range is [10,65535]. Default value is 65535.
 */
-LPFC_VPORT_ATTR_RW(tgt_queue_depth, 65535, 10, 65535,
-		   "Max number of FCP commands we can queue to a specific target port");
+static uint lpfc_tgt_queue_depth = LPFC_MAX_TGT_QDEPTH;
+module_param(lpfc_tgt_queue_depth, uint, 0444);
+MODULE_PARM_DESC(lpfc_tgt_queue_depth, "Set max Target queue depth");
+lpfc_vport_param_show(tgt_queue_depth);
+lpfc_vport_param_init(tgt_queue_depth, LPFC_MAX_TGT_QDEPTH,
+		      LPFC_MIN_TGT_QDEPTH, LPFC_MAX_TGT_QDEPTH);
+
+/**
+ * lpfc_tgt_queue_depth_store: Sets an attribute value.
+ * @phba: pointer the the adapter structure.
+ * @val: integer attribute value.
+ *
+ * Description: Sets the parameter to the new value.
+ *
+ * Returns:
+ * zero on success
+ * -EINVAL if val is invalid
+ */
+static int
+lpfc_tgt_queue_depth_set(struct lpfc_vport *vport, uint val)
+{
+	struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
+	struct lpfc_nodelist *ndlp;
+
+	if (!lpfc_rangecheck(val, LPFC_MIN_TGT_QDEPTH, LPFC_MAX_TGT_QDEPTH))
+		return -EINVAL;
+
+	if (val == vport->cfg_tgt_queue_depth)
+		return 0;
+
+	spin_lock_irq(shost->host_lock);
+	vport->cfg_tgt_queue_depth = val;
+
+	/* Next loop thru nodelist and change cmd_qdepth */
+	list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp)
+		ndlp->cmd_qdepth = vport->cfg_tgt_queue_depth;
+
+	spin_unlock_irq(shost->host_lock);
+	return 0;
+}
+
+lpfc_vport_param_store(tgt_queue_depth);
+static DEVICE_ATTR_RW(lpfc_tgt_queue_depth);
 
 /*
 # hba_queue_depth:  This parameter is used to limit the number of outstanding
diff --git a/drivers/scsi/lpfc/lpfc_debugfs.c b/drivers/scsi/lpfc/lpfc_debugfs.c
index fb0dc2aeed91..50c11acf73a8 100644
--- a/drivers/scsi/lpfc/lpfc_debugfs.c
+++ b/drivers/scsi/lpfc/lpfc_debugfs.c
@@ -544,7 +544,7 @@ static int
 lpfc_debugfs_nodelist_data(struct lpfc_vport *vport, char *buf, int size)
 {
 	int len = 0;
-	int cnt;
+	int i, iocnt, outio, cnt;
 	struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
 	struct lpfc_hba  *phba = vport->phba;
 	struct lpfc_nodelist *ndlp;
@@ -554,10 +554,12 @@ lpfc_debugfs_nodelist_data(struct lpfc_vport *vport, char *buf, int size)
 	struct nvme_fc_remote_port *nrport;
 
 	cnt = (LPFC_NODELIST_SIZE / LPFC_NODELIST_ENTRY_SIZE);
+	outio = 0;
 
 	len += snprintf(buf+len, size-len, "\nFCP Nodelist Entries ...\n");
 	spin_lock_irq(shost->host_lock);
 	list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
+		iocnt = 0;
 		if (!cnt) {
 			len +=  snprintf(buf+len, size-len,
 				"Missing Nodelist Entries\n");
@@ -585,9 +587,11 @@ lpfc_debugfs_nodelist_data(struct lpfc_vport *vport, char *buf, int size)
 			break;
 		case NLP_STE_UNMAPPED_NODE:
 			statep = "UNMAP ";
+			iocnt = 1;
 			break;
 		case NLP_STE_MAPPED_NODE:
 			statep = "MAPPED";
+			iocnt = 1;
 			break;
 		case NLP_STE_NPR_NODE:
 			statep = "NPR   ";
@@ -614,8 +618,10 @@ lpfc_debugfs_nodelist_data(struct lpfc_vport *vport, char *buf, int size)
 			len += snprintf(buf+len, size-len, "UNKNOWN_TYPE ");
 		if (ndlp->nlp_type & NLP_FC_NODE)
 			len += snprintf(buf+len, size-len, "FC_NODE ");
-		if (ndlp->nlp_type & NLP_FABRIC)
+		if (ndlp->nlp_type & NLP_FABRIC) {
 			len += snprintf(buf+len, size-len, "FABRIC ");
+			iocnt = 0;
+		}
 		if (ndlp->nlp_type & NLP_FCP_TARGET)
 			len += snprintf(buf+len, size-len, "FCP_TGT sid:%d ",
 				ndlp->nlp_sid);
@@ -632,10 +638,20 @@ lpfc_debugfs_nodelist_data(struct lpfc_vport *vport, char *buf, int size)
 			ndlp->nlp_usg_map);
 		len += snprintf(buf+len, size-len, "refcnt:%x",
 			kref_read(&ndlp->kref));
+		if (iocnt) {
+			i = atomic_read(&ndlp->cmd_pending);
+			len += snprintf(buf + len, size - len,
+					" OutIO:x%x Qdepth x%x",
+					i, ndlp->cmd_qdepth);
+			outio += i;
+		}
 		len +=  snprintf(buf+len, size-len, "\n");
 	}
 	spin_unlock_irq(shost->host_lock);
 
+	len += snprintf(buf + len, size - len,
+			"\nOutstanding IO x%x\n",  outio);
+
 	if (phba->nvmet_support && phba->targetport && (vport == phba->pport)) {
 		tgtp = (struct lpfc_nvmet_tgtport *)phba->targetport->private;
 		len += snprintf(buf + len, size - len,
diff --git a/drivers/scsi/lpfc/lpfc_scsi.c b/drivers/scsi/lpfc/lpfc_scsi.c
index 050f04418f5f..8570486013f3 100644
--- a/drivers/scsi/lpfc/lpfc_scsi.c
+++ b/drivers/scsi/lpfc/lpfc_scsi.c
@@ -3983,9 +3983,6 @@ lpfc_scsi_cmd_iocb_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pIocbIn,
 	}
 #endif
 
-	if (pnode && NLP_CHK_NODE_ACT(pnode))
-		atomic_dec(&pnode->cmd_pending);
-
 	if (lpfc_cmd->status) {
 		if (lpfc_cmd->status == IOSTAT_LOCAL_REJECT &&
 		    (lpfc_cmd->result & IOERR_DRVR_MASK))
@@ -4125,6 +4122,7 @@ lpfc_scsi_cmd_iocb_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pIocbIn,
 		msecs_to_jiffies(vport->cfg_max_scsicmpl_time))) {
 		spin_lock_irqsave(shost->host_lock, flags);
 		if (pnode && NLP_CHK_NODE_ACT(pnode)) {
+			atomic_dec(&pnode->cmd_pending);
 			if (pnode->cmd_qdepth >
 				atomic_read(&pnode->cmd_pending) &&
 				(atomic_read(&pnode->cmd_pending) >
@@ -4138,16 +4136,8 @@ lpfc_scsi_cmd_iocb_cmpl(struct lpfc_hba *phba, struct lpfc_iocbq *pIocbIn,
 		}
 		spin_unlock_irqrestore(shost->host_lock, flags);
 	} else if (pnode && NLP_CHK_NODE_ACT(pnode)) {
-		if ((pnode->cmd_qdepth != vport->cfg_tgt_queue_depth) &&
-		    time_after(jiffies, pnode->last_change_time +
-			      msecs_to_jiffies(LPFC_TGTQ_INTERVAL))) {
-			spin_lock_irqsave(shost->host_lock, flags);
-			pnode->cmd_qdepth = vport->cfg_tgt_queue_depth;
-			pnode->last_change_time = jiffies;
-			spin_unlock_irqrestore(shost->host_lock, flags);
-		}
+		atomic_dec(&pnode->cmd_pending);
 	}
-
 	lpfc_scsi_unprep_dma_buf(phba, lpfc_cmd);
 
 	spin_lock_irqsave(&phba->hbalock, flags);
@@ -4591,6 +4581,8 @@ lpfc_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *cmnd)
 				 ndlp->nlp_portname.u.wwn[7]);
 		goto out_tgt_busy;
 	}
+	atomic_inc(&ndlp->cmd_pending);
+
 	lpfc_cmd = lpfc_get_scsi_buf(phba, ndlp);
 	if (lpfc_cmd == NULL) {
 		lpfc_rampdown_queue_depth(phba);
@@ -4643,11 +4635,9 @@ lpfc_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *cmnd)
 
 	lpfc_scsi_prep_cmnd(vport, lpfc_cmd, ndlp);
 
-	atomic_inc(&ndlp->cmd_pending);
 	err = lpfc_sli_issue_iocb(phba, LPFC_FCP_RING,
 				  &lpfc_cmd->cur_iocbq, SLI_IOCB_RET_IOCB);
 	if (err) {
-		atomic_dec(&ndlp->cmd_pending);
 		lpfc_printf_vlog(vport, KERN_INFO, LOG_FCP,
 				 "3376 FCP could not issue IOCB err %x"
 				 "FCP cmd x%x <%d/%llu> "
@@ -4691,6 +4681,7 @@ lpfc_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *cmnd)
 	lpfc_scsi_unprep_dma_buf(phba, lpfc_cmd);
 	lpfc_release_scsi_buf(phba, lpfc_cmd);
  out_host_busy:
+	atomic_dec(&ndlp->cmd_pending);
 	return SCSI_MLQUEUE_HOST_BUSY;
 
  out_tgt_busy: